The study examines the literature, research, and data exploration on homicide patterns in the United States from 1976 to 2018. By merging yearly homicide crime data from the Federal Bureau of Investigation’s Supplementary Homicide Report, we show the trends of homicide patterns by year, state, age, and sex within the investigation period. In the analysis, regions in New York State, California, and Texas exhibit the highest homicide incidents. Among the offenders, the occurrence of the homicide incident has clustered between the age of 18 years old to 34 years old. In the U.S, male offenders composite 88% and female offenders composite 12% of all aggregated historical homicide incidents.
Keywords: Homicide, Crimes, Homicide Patterns, United States, Weapons
In 2020, the Federal Bureau of Investigation reported 21,570 total homicide cases in the United States(FBI, 2020). By analyzing the victim’s gender and ethnicity in the year 2020, researchers found the vast majority of victims by gender are African American males. William Julius Wilson’s seminal work, The Truly Disadvantaged, argued that racial difference in crime is due to inherent socio-economic circumstances. Wilson transformed his theory into the famous racial invariance hypothesis, which states that certain structural conditions disadvantaged various racial groups. In response to this hypothesis, researchers have focused on examining external factors on macro-economic, community circumstances, and other confounders contributing to the variable-specific outcome of homicide crimes through the FBI’s Uniform Crime Report.
The Federal Bureau of Investigation, the nation’s principal law enforcement agency, has compiled the Uniform Crime Report (UCR) to serve as the nation’s periodic assessment of reported crime in the U.S criminal justice system since the 1930s. Every year, the affiliated participating law enforcement agencies contribute reports directly or indirectly through their state programs. By the 1970s, the UCR data expanded to capture incident-specific data with the National Incident-Based Reporting System, or the NIBRS. In 1985, approximately 17,000 law enforcement agencies contributed to the report. In the 1990s, hate crime data are manifested in the collection of race, religion, sexual orientation, and ethnicity. In September 1994, disabilities, both physical and mental, were further added to the data. Following UCR’s development in reporting homicide incidents, researchers explored the data on homicide patterns in several publications.
Within criminology, for example, Steffensmeier et al. challenge the notion of ethnic violence bias based on poverty and inequality(Steffensmeier et al., 1992). The study echoes the theme from Edwin Sutherland (1947) on the perspectives of poverty and inequality as insufficient to explain racial crime differences. Instead, Steffensmeirr and el. explores the motives and behavioral framework that fosters homicide crime and crime-related incidents in the United States using the UCR data. Kreisman et al.(2022) explore juvenile homicides by analyzing the psychological factors of the offenders committing juvenile homicide crimes and the methodologies of crime prevention.
Allen et al. further investigate the idea of racial invariance in rural and urban contexts by comparing the structural factors of ethnicity, focusing on homicide among blacks, whites, and Latinx homicide from 2000 to 2010. (Allen et. al, 2022). Furthermore, gender is also a prevalent explanatory variable of data exploration by researchers. A study conducted by Asher et al(2022). focuses on a longitudinal study on the investigation of interpersonal among native Alaska women using the UCR data on reported homicide incidents.
Previous research has established the idea of Truly Disadvantaged contexts solely on the variety of external factors contributing to the crime. To understand the various factor contributing to homicide crime, we first aim to see the underlying patterns using historical evidence. In this study, we aim to explore the FBI’s UCR on homicide incidents in the United States from 1976 to 2018. The primary purpose is to visualize trends in homicide incidents by certain explanatory variables within the period of examination. The following section explains the methodology of the exploratory analysis.
The statistics of the dataset detail categorical variables on the offender’s information such as age, year, sex, race, state, state codes, reporting agency, the type of circumstance, and type of action by year from 1976 to 2018. The numeric variables include population and incident number. Additionally, the dataset provides details on the victim’s sex, age, race, ethnicity, the weapon used, the circumstance of the incident, and the relationship with the offender. Our analysis scope selected homicide data from 1976 to 2018 as it is an existing merged dataset published on ICPSR’s website.
In this study, we will use multivariable analysis applied to the time frame from 1976-2018 to explain the various crime patterns of homicide crimes over time. Specifically, we aim to explain the association between particular variables and the outcomes of homicide incidents in the United States. Given the magnitude of the dataset, we would first clean the unfilled information in the Jupyter notebook using python as the primary coding language.
The mega-dataset contains 159 columns and 72,1956 entries describing the homicide documentation by the offender’s county level, state, age, sex, weapons used, the relationship between the offender and the victim, and circumstances of the homicide. Given the magnitude of the data, our primary investigation is on categorical variables such as year, state, offender_age, offender_sex, and offender_race. We grouped seven variables and aggregated the number of total incidents in the United States to visualize the trends from 1976 to 2018. Specifically, we will aggregate the total number of homicide incidents and visualize the trends by time from the following variables.
We group by year and then aggregate the sum of incidents from the data entries. In the seaborn package in python, we will use a line plot to visualize the lowest and highest number of homicide incidents by year.
First, we drop the racial subgroup ‘p’ represented by the pacific island, which is statistically insignificant, and capitalize the strings in the data frame. We then group by race and aggregate the total incidents in the United States from 1976 to 2018. To calculate each race’s composition of homicide incidents, we would aggregate the total homicide incident and divide it by each ethnicity.
To visualize the homicide patterns by the 52 states in the United States, we first group by different states and then sum the total incidents that happened within the time frame of examination. In python, we use the Plotly-Express package to capture states by scale degrees.
Here, we clean and replace unfilled information such as ‘NaN,’ ‘Unknown,’ ‘bb,’ ‘nb.’ Since our age variable is categorical, we will convert it to numerical using the as.type(int) function in python. After cleaning the age variable, we assign offender age range from 1 to 110 years old into five categories: Under 18, 18-34, 35-49, ‘50-65’, and ‘65 or Older’. Then, we will assign each age to each bin and calculate which age group exhibits a high likelihood of committing homicide crime.
We clean the unfilled information such as ‘NaN’ and ‘Unknown.’ Therefore, the variable offender_1_sex is categorized as either male or female. In the Matplotlip package, we will plot the pie chart representing the homicide rate among each sex category.
Figure 1, the line plot represents the aggregated total homicide incident by year reported from 1976 to 2018. The peaks in the graph indicate that the U.S occurred the most homicide incidents in 1980 and 1993, with total incidents number of 15,849 incidents and 14,628 incidents, respectively. Since the early 2000s, the total homicide incidents have decreased, with the highest incidents in 2006 of 9628 cases. We suspect the clustered homicide incidents around 2006-2009 are associated with the macroeconomic environment in the unemployment rate, such as the global economic crisis.
To analyze the explanatory variable on race, we first implemented a time series visualization between the logarithmic homicide crime incidents by year among racial categories. As figure 2 shows, the number of incidents in the log is predominantly committed by white and black during the examination period. The race that committed the most negligible homicide crimes are Asians, American Indians, and Unknowns. Since the number of homicides committed by race varies substantially, we used the logistic function to normalize the differences. Through this method, we would better visualize the trends in crimes commited by different racial groups in the United States from 1976 to 2018.
The pie chart in figure 3 represents the racial composition in the percentage of all reported homicide incidents in the United States from 1976 to 2018. From the chart, we can see that the crimes committed by whites and blacks are relatively the same(49.1% vs. 47.9%). The highest percentage is whites(49.1%), Blacks(47.9%), and then followed by Asians(1.3%), American Indian or Alaska Native (0.9%), and Unknown(0.8%).
Figure 4 represents the total homicide incidents at the country level aggregated yearly from 1976 to 2018. States are categorized by levels of green, each level representing 10k. The darker the green, the more homicide incident was reported in that state. Some states have high levels of heterogeneity in total homicide incidents, including California and Texas, and New York. Most states have reported low homicide incidents, such as Wyoming and Vermont, which reported 382 and 670 incidents. Furthermore, Figure 4 also shows that most homicide incidents cluster in the eastern regions of the United States, such as New York, Georgia, and Illinois, reported 27,903, 17,750, and 19,739 incidents, respectively.
The histogram in figure 5 represents the total homicide incidents from 1976 to 2018 in different states, ranked by the highest and the lowest incident states from right to left. The highest homicide incident reported states are California, Texas, and New York, which reported 70,405, 52,032, and 27,903 incidents, respectively. The lowest homicide incident reported states are Vermont, North Dakota, and Guam, each representing 382, 358, and 19 incidents.
After demonstrating the spatial distribution of homicide incidents by state, we now analyze the homicides incident by age. On the x-axis, we represent each age group as ‘Under 18’, ‘18-34’, ‘35-49’, ‘50-65’, and ‘65 or Older. From figure 6, we can see that most offenders in homicide incidents are between 18 years old to 34 years old, followed by offenders between 35 to 49 years old, under 18 years old, between 50-65 years old, and 65 or older.
The offenders’ sex, categorized by males and females, is the final explanatory variable of our examination. Figure 7 represents the sex composition of the offenders from 1976 to 2018 and 1995. In the first pie chart, we show that male offenders composite 88% of homicide incidents reported from 1976-2018, while females only represent 12%. Of the reported incidents, 42,6207 incidents were male, and 57,971 incidents were female. In 1995, male offenders represented 90.4% of total homicide incidents, while female offenders only represented 9.6%. We chose the year 1995 as an examination because it is a year that reported a high number of homicide incidents report. Therefore, we aim to see if there’s a significant difference between male and female offenders.
Allen, Caroline B., and Ben Feldmeyer. “Racial Invariance in Rural and Urban Contexts: Comparing the Structural Sources of Black, White, and Latinx Homicide in Rural and Urban Counties, 2000 and 2010.” Taylor & Francis Online: Peer-Reviewed Journals
Asher, Autumn. “Protecting Native Motherhood: A Longitudinal Investigation of Interpersonal Violence Among Alaska Native Women.” Washington University in St.Louis, Arts & Sciences Electronic Theses and Dissertations
Jennings, James. “Book Review: The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy by William Julius Wilson.” ScholarWorks at UMass Boston, 22 Sept. 1988
Kreisman J Kreisman; R Seiden, J, and R Seiden. “Juvenile Homicide - A Literature Review.” Juvenile Homicide - A Literature Review. Office of Justice Programs, 13 Jan. 1984
Steffensmeier, Darrell. “Scope and Conceptual Issues in Testing the Race-Crime Invariance Thesis: Black, White, and Hispanic Comparisons.” Wiley Online Library, 28 Nov. 2010
Federal Bureau of Investigation. “Uniform Crime Reporting Program Resource Guide. NACJD.” Uniform Crime Reporting Program Resource Guide
[NbConvertApp] Converting notebook Crime1976_2018_new.ipynb to html [NbConvertApp] Writing 4559502 bytes to Crime1976_2018_new.html
[NbConvertApp] Converting notebook Crime1976_2018_new.ipynb to html [NbConvertApp] Writing 4695300 bytes to Crime1976_2018_new.html